Network Averaging: a technique for determining a proxy for the dynamics of networks

The main aim of this paper is to introduce the network averaging technique. This technique is introduced because accurately determining the structure of real networks can be difficult and the network averaging technique provides a proxy for real networks. A second aim is to introduce the adaptive interactive expectations (AIE) model, which uses a ‘pressure to change profit expectations index’ to replace the utility curve maximising agent concept. The AIE model has an interactive expectations network, which is difficult to determine, so suitable to illustrate network averaging. The AIE model is tested against the Dun and Bradstreet Profit Expectations Survey. The paper finds network averaging improves the predictive performance of AIE over its benchmarks: the rational expectations hypothesis and the adaptive expectations model. The network averaging technique could be adapted to other situations where there are endogenous effects acting through difficult to measure networks. The AIE model could be readily applied to other forms of expectations and as a replacement for the utility curve maximising agent. Finally, in this paper AIE models profit expectations, which are an important issue in their own right because they affect investment decisions and whether one business will extend credit to another business.